[Ffmpeg-devel-irc] ffmpeg-devel.log.20160930

[burek]

[00:24:18 CEST] <llogan> michaelni__: did you unsubscribe the bouncing user that cehoyos mentioned? that address is not in the membership list. could be an alias...
[00:26:04 CEST] <michaelni__> llogan, i wanted to look into it but forgot, so no i didnt unsub him
[00:28:47 CEST] <llogan> i guess I'll email the user directly and ask him what the alias is
[00:30:21 CEST] <llogan> ...becasue fixing DKIM seems like too much of a pain in the ass
[00:34:02 CEST] <wm4> I already regret "returning" to ffmpeg development
[00:34:05 CEST] <wm4> all this ridiculous crap...
[00:34:28 CEST] <wm4> I can only hope project management will be different in the future
[00:43:08 CEST] <jamrial> wm4: it's not the first time a patch that makes such a big change breaks a weird file or scenario
[00:43:29 CEST] <jamrial> the codecpar and bsf ports also broke a lot of things at first
[00:47:14 CEST] <wm4> that's not the problem, but that it has to be libavdevice/openal
[00:56:22 CEST] <iive> wm4: what's the problem?
[00:58:48 CEST] <wm4> bad test case
[01:14:34 CEST] <iive> ??
[01:15:21 CEST] <Chloe> iive: wm4 doesn't want to use openal, so they want a different way to produce the issue michaelni is describing (not involving lavd)
[01:16:03 CEST] <iive> what's the issue?
[01:16:37 CEST] <Chloe> The testcase uses openal
[01:17:09 CEST] <iive> testcase for what?
[01:18:15 CEST] <llogan> refer to the ML
[01:18:32 CEST] <Chloe> iive: https://ffmpeg.org/pipermail/ffmpeg-devel/2016-September/200319.html
[07:56:11 CEST] <Wesley_Chan> Hi.
[08:00:35 CEST] <Wesley_Chan> I'm not sure if anyone is listening to this but I want "-map_metadata 0" to copy nothing but only the language part. I had to modify av_dict_copy() in libavutil/dict.c to archieve this. I know this is a dirty fix but is there a better place to put my patch? Maybe a "-metadata_cleanup" switch in the future?
[08:34:56 CEST] <rcombs> hmm, I've been considering a similar feature
[08:35:00 CEST] <rcombs> too bad they're gone
[08:35:09 CEST] <rcombs> might do it with a tweak to copy_metadata to let you specify specific keys to map
[09:21:27 CEST] <Chloe> michaelni: I still don't understand, why couldn't the libfaac testcases be moved to another encoder?
[09:23:03 CEST] <nevcairiel> there arent even actual proper testcases, not in the sense of FATE
[09:23:12 CEST] <nevcairiel> because we dont test external libnraries
[10:03:44 CEST] <wm4> michaelni: do you really expect me to install openal and to enable libavdevice?
[10:05:23 CEST] <ubitux> how portable is regex.h?
[10:07:02 CEST] <rcombs> ubitux: probably portable enough to have as an optional dep, but not portable enough to require for anything core?
[10:07:13 CEST] <rcombs> (available on win32, but a pain)
[10:07:20 CEST] <ubitux> my guess as well, ok
[10:07:24 CEST] <rcombs> what for?
[10:07:35 CEST] <ubitux> there is a patch for multistream SAMI 
[10:07:52 CEST] <ubitux> that format needs you to parse the CSS style to extract languages specifications (don't ask me.)
[10:08:05 CEST] <ubitux> the patch is just doing some regex to find them
[10:09:28 CEST] <rcombs> sounds like it should probably be rewritten without regex, but also wow what a mess
[10:09:34 CEST] <ubitux> yeah
[10:09:41 CEST] <ubitux> it's not the right approach too
[10:09:45 CEST] <ubitux> it needs to create multiple streams
[10:12:15 CEST] <wm4> ubitux: it's not?
[10:12:37 CEST] <wm4> (haven't looked at the patch)
[10:12:58 CEST] <ubitux> it's not creating multiple streams for each language
[10:14:11 CEST] <rcombs> would be nice to have some mechanism for a decoder (parser?) to split input into multiple streams
[10:14:32 CEST] <rcombs> happens for e.g. subtitles crammed into H.264, too
[10:15:59 CEST] <wm4> well this is a demuxer
[10:16:06 CEST] <wm4> or should be
[10:30:45 CEST] <rcombs> wm4: so, pass streams coming from e.g. a .ts through a "chained" demuxer?
[10:31:19 CEST] <wm4> I meant the sami thing
[10:31:38 CEST] <rcombs> oh, yeah, if that's in a plaintext sami file sure
[10:32:00 CEST] <rcombs> for some reason I thought we were talking about multiple actual streams in the same e.g. MKV track
[10:32:02 CEST] <wm4> for A53 captions, feeding the side data back to a pseudo-demuxer stream might work (mpv does that)#
[10:32:32 CEST] <rcombs> I'd just like to see it handled within lavf
[10:33:49 CEST] <wm4> for that it'd have to be able to do frame reordering within the parser
[11:51:22 CEST] <michaelni> Chloe, to test a ticket for regression one looks at the ticket on trac, takes the input sample and the quoted command line from the ticket. Now to truly move these to a different encoder one would have to check if the original issue is reproducable with a diferent encoder with old FFmpeg and if needed adjust the testcase until it reproduces then update the testcase in the ticket. Noone will do this, its a completely 
[11:51:22 CEST] <michaelni> unrealistic amount of work 
[11:54:59 CEST] <nevcairiel> keeping some old piece of unused code for this a lone seems also rather silly
[13:13:22 CEST] <rcombs> https://puu.sh/rsOdK/e9eca8a945.ts mmmmmmm update_stream_timings issues
[13:14:34 CEST] <rcombs> https://gist.github.com/50d4b9123bae598d9c3d1a9f321e44b7 fixes
[13:25:15 CEST] <jsodjfcn> am I readable ?
[13:26:02 CEST] <jsodjfcn> am I readable ? any developers around ? wanted to submit a bug if its a bug ?
[13:27:13 CEST] <wm4> jsodjfcn: you are readable
[13:27:26 CEST] <wm4> rcombs: looks scary
[13:27:41 CEST] <BtbN> "git: 'send-email' is not a git command. See 'git --help'." I'm quite sure it is git.
[13:27:43 CEST] <rcombs> wm4: there's already almost identical code for end times
[13:28:18 CEST] <wm4> BtbN: you probably need to install it separately
[13:28:39 CEST] <jsodjfcn> wm4: are you a devel ? should I start my bug story ?
[13:29:00 CEST] <wm4> jsodjfcn: bug reports go on the bug tracker
[13:29:11 CEST] <jsodjfcn> wm4: I dont know if its a bug
[13:29:27 CEST] <wm4> normally you'd go to #ffmpeg or so and ask there
[13:29:45 CEST] <wm4> (let's just say they have better experience dealing with user problems)
[13:29:55 CEST] <jsodjfcn> wm4: I'm coming from there only, there was only BtbN responding and helping out
[13:30:01 CEST] <wm4> ubitux: is there a chance yet to have a new AVFrame based subtitle API?
[13:30:21 CEST] <nevcairiel> how do you put text into AVFrames
[13:30:30 CEST] <wm4> that's the question
[13:31:22 CEST] <nevcairiel> unifying subtitles into the same structures would definitely be great
[13:31:31 CEST] <BtbN> wm4, indeed. Seems to be a new feature.
[13:31:35 CEST] <wm4> yeah, just for the APIs
[13:32:10 CEST] <nevcairiel> but then, subtitles have some additional metadata that AVFrames just dont carry yet
[13:32:18 CEST] <nevcairiel> like, position
[13:32:40 CEST] <wm4> if anything the AVFrame could merely point to an AVSubtitle
[13:33:06 CEST] <ubitux> wm4: yeah, now that i'm done with nlmeans and codecpar, i have time for this
[13:33:38 CEST] <wm4> ubitux: once this is done, ffmpeg.c's data flow could possibly be changed to something that's nicer with the new APIs
[13:33:38 CEST] <ubitux> i'm a bit anxious about the libav merge commits stack though
[13:33:46 CEST] <ubitux> wm4: yeah, i know ;)
[13:33:54 CEST] <ubitux> that will be grea
[13:33:56 CEST] <ubitux> +t
[13:34:15 CEST] <wm4> also, does ffprobe have any tests?
[13:34:17 CEST] <nevcairiel> you guys got a lot of the "bad" backlog of merges out of the way
[13:34:20 CEST] <nevcairiel> it does
[13:34:31 CEST] <wm4> within the fate target?
[13:34:34 CEST] <nevcairiel> yes
[13:34:37 CEST] <wm4> oh, ok
[13:34:53 CEST] <rcombs> wm4: yeah, fate-ffprobe-<format>
[13:35:17 CEST] <rcombs> and a fair number of other tests are based on comparison of ffprobe output
[13:35:22 CEST] <nevcairiel> not sure what other things it might be used for in the tests
[13:38:06 CEST] <jsodjfcn> wm4: 1st part of story- I download a video of 2 parts from web and joined them with ffmpeg *** 2nd part- joinging with latest version from zerone plays only 1st part and not 2nd part video, audio however plays fine. **** 3rd part - joining the same 2 vids with an older version of ffmpeg embedded with thekmplayer plays video fine. both parts and audio also. *** now is it some kind of bug in new version ?
[13:38:26 CEST] <wm4> that's definitely a #ffmpeg question, go there
[13:38:50 CEST] <jsodjfcn> wm4: told you, I'm coming from there only
[13:39:02 CEST] <wm4> then wait longer for a reply
[13:39:09 CEST] <wm4> this channel is not for user help
[13:39:10 CEST] <BtbN> and you already got the right reply. Please report it on trac.
[13:40:59 CEST] <jsodjfcn> I've already joined video with older version, I need no further help on this particular matter, I just wanna help ffmpeg in bringing to its notice a bug if its a bug
[13:41:31 CEST] <wm4> then report it on the bug tracker
[13:42:08 CEST] <jsodjfcn> BtbN: thnx for your help, but reporting a bug doesnt completes a task currently on hand
[13:42:23 CEST] <jsodjfcn> anyways thnx all, BtbN thnx
[13:42:36 CEST] <BtbN> by the time you have spent arguing about this you could have filed the bug multiple times.
[13:43:20 CEST] <jsodjfcn> BtbN: I'm not arguing, just clearing my point, I was very very clear, I dont wanna create a new account
[13:47:19 CEST] <Compn> jsodjfcn : whats the url of your files and i'll post the trac bug for you
[13:47:29 CEST] <Compn> also need full output from both ffmpeg versions
[13:47:46 CEST] <Compn> including the commands you used
[13:47:57 CEST] <Compn> paste it all to pastebin so i can grab it easily...
[13:48:48 CEST] <Compn> without your specific sample files, its possible we wont find the bug 
[13:49:59 CEST] <Compn> oh also isnt there pre made account people can use on ffmpeg trac to report bugs ? 
[13:50:03 CEST] <Compn> hmm theres one for mplayer ;D
[13:50:26 CEST] <Compn> that would make it easier for people who dont want to register to post bugs :)
[13:51:01 CEST] <jsodjfcn> Compn: okay, gimme a few minutes, I'll do it slowly, I'm not a pro (I support FFMPEG :D)
[13:51:09 CEST] <Compn> thanks :)
[14:13:37 CEST] Action: Compn goes kinda afk will read later
[15:41:06 CEST] <BBB> atomnuker: isnt faac an encoder?
[15:41:35 CEST] <BBB> (I think it was just a typo, you seem to be talking about encoding stuff in yuor response, but you said decoder instead of en)
[15:42:26 CEST] <wm4> yeah it's only an encoder
[15:43:02 CEST] <wm4> man I remember mplayer times, and how I sometimes had to switch decoder libs for various broken streams or decoder bugs
[15:43:08 CEST] <wm4> usually for aac
[15:43:45 CEST] <wm4> libfaad
[15:44:34 CEST] <wm4> wow there's even libfaad2
[15:44:39 CEST] <wm4> which gstreamer uses (???)
[15:45:07 CEST] <wm4> trying to remove it from debian will literally wipe all KDE packages
[15:45:09 CEST] <wm4> so backwards
[15:45:49 CEST] <atomnuker> wm4: shouldn't do it if you have the null phonon backend
[15:46:30 CEST] <atomnuker> (though dependencies are messed up wrt phonon backends and its hard not to install vlc)
[16:01:45 CEST] <BBB> linux package management...
[16:01:49 CEST] <BBB> omg, the memories
[16:02:45 CEST] <rcombs> at least it's not windows
[16:02:49 CEST] <wm4> or OSX
[16:03:07 CEST] <BBB> and the trollwar began :D
[16:03:10 CEST] Action: BBB runs
[16:24:55 CEST] <philipl> What happens with hwframes if the frame size changes mid-stream? I don't see it getting reinitialized, but maybe I'm missing something.
[16:31:38 CEST] <nevcairiel> hwaccels will typically just re-create the entire decoder block, with new hwframes
[16:31:47 CEST] <nevcairiel> non-hwaccel t hings will probably have to die horribly
[16:39:32 CEST] <philipl> ok. So I'm reading it correctly :-)
[19:41:00 CEST] <ubitux> http_multiclient crashes out of the box
[19:41:14 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:44103c600167: build: sort EXAMPLE_LIST
[19:41:14 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:352ca77f8ab5: build: fix avio_dir_cmd and avio_reading example dependencies
[19:41:14 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:fb5e58cd773e: doc/examples/http_multiclient: fix compilation
[19:41:14 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:f2579f7ee232: doc/examples/http_multiclient: fix mixed declarations and code
[19:41:14 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:3d064b7802fd: build: remove references to inexistant avcodec example
[19:41:15 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:1e647d200fae: doc/examples/gitignore: add forgotten qsvdec entry
[19:41:15 CEST] <cone-427> ffmpeg 03Clément BSsch 07master:1a9b4bc4c7b6: doc/examples: build http_multiclient example
[19:41:41 CEST] <ubitux> http://sprunge.us/NiQE
[19:41:59 CEST] <ubitux> anyone familiar with this?
[20:15:25 CEST] <BBB> ubitux: looks like a NULL pointer being used somewhere
[20:15:28 CEST] <BBB> no I havent used it before
[20:24:40 CEST] <Chloe> ubitux: nice prompt
[20:25:15 CEST] <ubitux> Chloe: and you don't even see the   on failure ;)
[20:25:53 CEST] <Chloe> oh wow, I just have a '$' which change colours, your prompt sounds really fancy
[23:47:34 CEST] <ubitux> Chloe: not really; http://sprunge.us/gDAD
[23:48:23 CEST] <ubitux> (+setopt transient_rprompt)
[23:48:36 CEST] <Chloe> is that just bash?
[23:48:45 CEST] <ubitux> zsh
[23:50:10 CEST] <Chloe> ah, maybe I should add some emoji to mine as well. my zsh prompt: PS1="%6F%n%f@%5F%m%f:%4F%~%f%(!.%(?.%2F#%f.%1F#%f).%(?.%2F$%f.%1F$%f)) " I made it when I understood how it worked, I should have probably used the long options instead of the short ones
[00:00:00 CEST] --- Sat Oct  1 2016